Introduction

The volcano3D package enables exploration of probes differentially expressed between three groups. Its main purpose is for the visualisation of differentially expressed genes in a three-dimensional volcano plot. These plots can be converted to interactive visualisations using plotly.

This vignette consists of a case study from the PEAC rheumatoid arthritis project (Pathobiology of Early Arthritis Cohort). The methodology has been published in ‘Lewis, Myles J., et al. “Molecular portraits of early rheumatoid arthritis identify clinical and treatment response phenotypes.” Cell reports 28.9 (2019): 2455-2470.’ (DOI: 10.1016/j.celrep.2019.07.091) with an interactive web tool available at https://peac.hpc.qmul.ac.uk.

Getting Started

Prerequisites

Install from CRAN

Not yet publicly available:

install.packages("volcano3D")

Install from Github

library(devtools)
install_github("KatrionaGoldmann/volcano3D")

library(volcano3D)

Sample Data

The sample data used in this vignette can be loaded from the volcano3Ddata package. This is only possible after volcano3D is imported.

install.packages("volcano3Ddata")

Dictionary

Variables used in this vignette:

Variable	Definition
contrast	the variable by which samples can be split into three groups.
groups	the three levels/categories of the contrast variable.
comparison	two groups between which a statistical test can be performed. There are three comparisons total. For the examples outlined in this vignette we look at comparisons: ‘lymphoid-myeloid’, ‘lymphoid-fibroid’ and ‘myeloid-fibroid’.
p	p value
FC	fold change
padj	adjusted p value
suffix	the tail word in a column name. In this package it states the statistical parameter (e.g. _logFC is the log FC variable)
prefix	the leading word in a column name. In this package it states the statistical test (e.g. LRT is the likelihood ratio test).
polar	A polar coordinates object, of S4 class, containing the expression data, sample data, pvalues and polar coordinates.

Examples

This vignette will demonstrate the power of this package using a basic example from the PEAC data set. Here we will focus on the synovial data from this cohort.

Synovial Gene Data

First we will set up a polar object, using the polar_coords function, which contains:

the expression data with columns representing different samples and rows representing different variables
the sample data which contains information for each sample in rows.
the pvalues which contains the statistical significance of probes between groups.

To create the polar coordinates you must provide:

Dataframe	Column(s)	Purpose
sampledata	ID	Contains the sample IDs. This must be titled ‘ID’.
	contrast	A column containing the three-level factor used for contrasts. This column name must match the contrast variable
pvalues	pvalue	Three pvalues columns with the differential expression between groups. Column names are of format `paste0(groups[i], "-", groups[j], " ", p_col_suffix)`
	fold change (optional, x3)	Three optional fold change columns. One for each comparison between groups. Column names are of format `paste0(groups[i], "-", groups[j], " ", fc_col_suffix)`.
	adjusted pvalue (optional, x3)	Three optional adjusted pvalue columns. One for each comparison between groups. Column names are of format `paste0(groups[i], "-", groups[j], " ", padj_col_suffix)`.
	multi-group pvalue (optional)	An optional multi-group pvalue column. This is typically generated using ANOVA or likelihood ratio tests between all three groups. Column names of form `paste(multi_group_prefix, p_col_suffix)`
	multi-group fold change (optional)	An optional multi-group fold change column. This is typically generated using ANOVA or likelihood ratio tests between all three groups. Column names of form `paste(multi_group_prefix, fc_col_suffix)`
	multi-group adjusted pvalue (optional)	An optional multi-group adjusted pvalue column. This is typically generated using ANOVA or likelihood ratio tests between all three groups. Column names of form `paste(multi_group_prefix, padj_col_suffix)`
	label (optional)	An optional character column with probe labels for downstream plotting.

with additional arguments:

Variable	Purpose
contrast	The column name in sampledata which contains the variable by which samples can be split into three groups.
groups (optional)	The groups to be compared (in order). If NULL this defaults to levels(sampledata[, ‘contrasts’])`.
p_col_suffix (optional)	The suffix of column names with pvalues (default is ‘pvalue’).
padj_col_suffix (optional)	The suffix of column names with adjusted pvalues (defaults ‘padj’). If NULL the adjusted pvalue is calculated using p_col_suffix and padjust_method.
padjust_method (optional)	The method to calculate adjusted pvalues if not already provided. Must be one of c(‘holm’, ‘hochberg’, ‘hommel’, ‘bonferroni’, ‘BH’, BY’, ‘fdr’, ‘none’). Default is ‘BH’.
fc_col_suffix (optional)	The suffix of column names with log(fold change) values default is ‘logFC’).
multi_group_prefix (optional)	The prefix for columns containing statistics for a multi-group test (this is typically a likelihood ratio test or ANOVA). Default is NULL.
label_column (optional)	A column name in pvalues which is to be used to label markers of interest at plotting stage. If NULL the rownames will be used.

A vignette with examples to calculate the pvalues dataframes can be found in the pvalues vignette.

Using the synovial biopsies from PEAC we can create a polar object for differentially expressed genes. First we will install the volcano3Ddata package if necessary:

install.packages("volcano3Ddata")

Then the data can be loaded and polar coordinates calculated:

library(volcano3Ddata)
data("syn_data")

syn_polar <- polar_coords(sampledata = syn_metadata,
                          contrast = "Pathotype",
                          pvalues = syn_pvalues,
                          expression = syn_rld,
                          p_col_suffix = "pvalue",
                          padj_col_suffix = "padj",
                          fc_col_suffix = "log2FoldChange",
                          multi_group_prefix = "LRT",
                          non_sig_name = "Not Significant",
                          significance_cutoff = 0.01,
                          label_column = NULL,
                          fc_cutoff = 0.1)

The pvalues slot should have at least two statistics for each comparison - pvalue and adjusted pvalue with an optional logarithmic fold change statistic also:

head(syn_polar@pvalues)

	Fibroid-Lymphoid logFC	Fibroid-Lymphoid padj	Fibroid-Lymphoid pvalue	LRT logFC	LRT padj	LRT pvalue	Lymphoid-Myeloid logFC	Lymphoid-Myeloid padj	Lymphoid-Myeloid pvalue	Myeloid-Fibroid logFC	Myeloid-Fibroid padj	Myeloid-Fibroid pvalue	label
A2M	-0.0478326	1	0.7775361	0.2681094	0.4353436	0.2739571	-0.2202768	1.0000000	0.1563045	0.2681094	1	0.1694654	A2M
A2ML1	1.6854024	1	0.0002106	0.2445664	0.0000202	0.0000022	-1.9299687	0.0464724	0.0000030	0.2445664	1	0.6348844	A2ML1
A4GALT	1.0248506	0	0.0000000	-0.4601841	0.0000000	0.0000000	-0.5646665	0.2927492	0.0000192	-0.4601841	1	0.0055222	A4GALT
A4GNT	-1.1527859	1	0.0238302	1.3385294	0.1583022	0.0732984	-0.1857435	1.0000000	0.6691773	1.3385294	1	0.0197330	A4GNT
AAAS	-0.0245296	1	0.8574281	-0.0303402	0.9711813	0.9075226	0.0548697	1.0000000	0.6608837	-0.0303402	1	0.8469982	AAAS
AACS	0.1729592	1	0.2815302	0.1002448	0.2732939	0.1472059	-0.2732040	1.0000000	0.0630182	0.1002448	1	0.5874802	AACS

The sig column in syn_polar@polar allows us to determine relative differences in expression between groups (in this case pathotypes). The ‘+’ indicates which pathotypes are significantly ‘up’ compared to others. For example:

genes labelled ‘Lymphoid+’ are significantly up in Lymphoid vs Myeloid and Lymphoid vs Fibroid.
genes up in two pathotypes such as ‘Lymphoid+Myeloid+’ are up in both Lymphoid and Myeloid, therefore Lymphoid vs Fibroid and Myeloid vs Fibroid are statistically significant.
genes which show no significant difference between pathotypes are classed according to non_sig_name

This gives us:

setNames(data.frame(table(syn_polar@polar$sig)), c("Significance", "Frequency"))

Significance	Frequency
Fibroid+	885
Fibroid+Lymphoid+	19
Fibroid+Myeloid+	1504
Lymphoid+	1793
Lymphoid+Myeloid+	500
Myeloid+	119
Not Significant	11415

Volcano Plots

If there is a fold change column previously provided, we can now investigate the comparisons between pathotypes using the volcano_trio function. This creates three ggplot outputs.

syn_plots <- volcano_trio(polar = syn_polar,
                          sig_names = c("not significant","significant",
                                        "not significant","significant"),
                          colours = rep(c("grey60",  "salmon"), 2),
                          text_size = 9,
                          shared_legend_size = 0.9,
                          label_rows = c("SLAMF6", "PARP16", "ITM2C"),
                          fc_line = FALSE)

syn_plots$All

Radial Plots

These can be output to an interactive radar plot using radial_plotly. The labelRows variable allows any markers of interest to be labelled.

radial_plotly(polar = syn_polar, label_rows = c("SLAMF6", "PARP16", "ITM2C"))

By hovering over certain point you can also determine genes for future interrogation.

Similarly we can create a static ggplot image using radial_ggplot. We can also convert to a continuous colour scale by converting the angle to a hsv variable. This angle can be offset by continuous_shift.

radial_ggplot(polar = syn_polar,
              label_rows = c("SLAMF6", "PARP16", "ITM2C"),
              marker_size = 2.3,
              marker_alpha=0.3,
              colour_scale = "continuous",
              continuous_shift=120)

Boxplots

We can then interrogate any one specific variable as a boxplot, to investigate these differences. This is build using ggplot2 so can easily be edited by the user to add features.

plot1 <- boxplot_trio(syn_polar,
                      value = "SLAMF6",
                      test = "polar_padj",
                      levels_order = c("Lymphoid", "Myeloid", "Fibroid"),
                      box_colours = c("blue", "red", "green3"),
                      step_increase = 0.1) +
  labs(title="Adjusted Pvalues")

plot2 <- boxplot_trio(syn_polar,
                      value = "SLAMF6",
                      test = "polar_multi_padj",
                      levels_order = c("Lymphoid", "Myeloid", "Fibroid"),
                      box_colours = c("blue", "red", "green3")) +
  labs(title="Overall Adjusted Pvalue")

plot3 <- boxplot_trio(syn_polar,
                      value = "PARP16",
                      test = "t.test",
                      levels_order = c("Myeloid", "Fibroid"),
                      box_colours = c("red", "green3")) +
  labs(title="T-test")

ggarrange(plot1, plot2, plot3, ncol = 3)

Three Dimensional Volcano Plots

The final thing we can look at is the 3D volcano plot which projects differential gene expression onto cylindrical coordinates.

p <- volcano3D(syn_polar,
               label_rows = c("SLAMF6", "PARP16", "ITM2C"),
               label_size = 10,
               xy_aspectratio = 1,
               z_aspectratio = 0.9)

p %>% layout(legend = list(x = 100, y = 0.5))

Again this produces an interactive plot. If you have the orca command-line utility installed, this can be used to save static images. To install follow the instructions here.

orca(p, "./volcano_3d_synovium.svg", format = "svg")

Altering the Grid

By default volcano3D generates a grid using 12 spokes. If you wish to override any of these variables it is possible to pass in your own grid. This is also possible for the 3D radial plots where the z elements can be left as NULL.

By manually creating a grid object it is possible to change the tick points on axes as well as the number of radial spokes (n_spokes).

The default ticks are calculated from r_vector and z_vector using the pretty function. This can be overwritten by passing tick points in r_axis_ticks and z_axis_ticks.

For example we can decrease the number of radial spokes to 4, while altering the z axis ticks:

four_grid = polar_grid(r_vector=syn_polar@polar$r_zscore,
                       z_vector=NULL,
                       r_axis_ticks = NULL,
                       z_axis_ticks = c(0, 8, 16, 32),
                       n_spokes = 4)

similarly to extend the radial axis and increasing the number of spokes in 2D we can apply:

new_grid = polar_grid(r_vector=NULL,
                        z_vector=-log10(syn_polar@pvalues$`LRT pvalue`),
                        r_axis_ticks = c(1, 2, 3),
                        z_axis_ticks = NULL,
                        n_spokes = 24)

Altering the Labels

By default the rownames of the pvalues object is used to label markers of interest. This can be altered by amending the label column in the polar object:

syn_polar@polar$label <- paste0(syn_polar@polar$label, "!")

radial_ggplot(polar = syn_polar,
              label_rows = c("SLAMF6", "PARP16", "ITM2C"),
              grid = new_grid,
              marker_size = 2.3,
              legend_size = 10) + 
  theme(legend.position=c(1.1, 1))

Citation

If you use this package please cite as:

Lewis, Myles J., et al. “Molecular portraits of early rheumatoid arthritis identify clinical and treatment response phenotypes.” Cell reports 28.9 (2019): 2455-2470.

or using:

citation("volcano3D")

## 
## To cite package 'volcano3D' in publications use:
## 
##   Katriona Goldmann and Myles Lewis (2020). volcano3D: Interactive
##   Plotting of Three-Way Differential Expression Analysis. R package
##   version 1.0.0.9000. https://github.com/KatrionaGoldmann/volcano3D
## 
## A BibTeX entry for LaTeX users is
## 
##   @Manual{,
##     title = {volcano3D: Interactive Plotting of Three-Way Differential Expression
## Analysis},
##     author = {Katriona Goldmann and Myles Lewis},
##     year = {2020},
##     note = {R package version 1.0.0.9000},
##     url = {https://github.com/KatrionaGoldmann/volcano3D},
##   }

volcano3D Vignette

Katriona Goldmann